- Number
- 20260330257
- Published
- 2026-10-22
- Filed
- 2026-06-29
- Assignee
- Disney Enterprises, Inc.
- Inventors
- Fidaleo; Douglas A., Kennedy; James R., Barron; Michael
- CPC
- G06N5/022
- Verdict
- Low Notable software
- First reported
- 2026-W43 (2026-10-23)
- Source
- Google Patents · FreePatentsOnline
The keeper's note
An interactive AI character asks a guest to share a memory, scores how closely it matches memories it already holds, and shifts its emotional response when nothing matches well.
Abstract
There is provided a method including eliciting, via an artificial intelligence interactive character (AIIC), a reminiscence from a user, predicting, using a trained machine learning (ML) model, at least one user memory feature from the reminiscence, computing, for each memory node of the plurality of memory nodes, a bond factor between the predicted at least one user memory feature and the one or more memory features stored in the memory node to provide a plurality of bond factors, determining that none of the plurality of bond factors satisfies a threshold, and in response to determining that none of the plurality of bond factors satisfies the threshold: updating the memory data structure by inserting a new memory node storing the predicted at least one user memory feature in the memory data structure; and generating, using the updated memory data structure, an interactive communication for execution by the AIIC.
Background
BACKGROUND
Establishing a deep emotional connection between a human and an artificial intelligence (AI) character is an unsolved problem of significant importance to many fields. Although there is evidence in the psychology literature that sharing memories creates a sense of relationship closeness between individuals, that apparent ability to share memories has not heretofore been extended to AI characters. The emotional closeness engendered by the sharing of memories is often enhanced when these memories are relatable between the individuals, i.e., the individuals have similar experiences or interpretations of those experiences. Such closeness generally improves communication between individuals and tends to make interactions richer and more pleasurable. However, until now, AI agents have had only crude ability to mimic human emotional behavior, which may be off-putting rather than enjoyable. Thus, there exists a need in the art for systems and methods to improve the ability of AI characters to express language and behaviors in a manner similar to individuals having shared memories.
Claims
1. A system comprising: a computing platform comprising a hardware processor and a system memory; the system memory storing software code and a memory data structure comprising a plurality of memory nodes, each memory node of the plurality of memory nodes storing one or more memory features for an artificial intelligence interactive character (AIIC); and the hardware processor configured to execute the software code to: elicit, via the AIIC, a reminiscence from a user; predict, using a trained machine learning (ML) model, at least one user memory feature from the reminiscence; compute, for each memory node of the plurality of memory nodes, a bond factor between the predicted at least one user memory feature and the one or more memory features stored in the memory node to provide a plurality of bond factors; determine that none of the plurality of bond factors satisfies a threshold; and in response to determining that none of the plurality of bond factors satisfies the threshold: update the memory data structure by inserting a new memory node storing the predicted at least one user memory feature in the memory data structure; and generate, using the updated memory data structure, an interactive communication for execution by the AIIC. ||
13. A method comprising: eliciting, via an artificial intelligence interactive character (AIIC), a reminiscence from a user; predicting, using a trained machine learning (ML) model, at least one user memory feature from the reminiscence; computing, for each memory node of the plurality of memory nodes, a bond factor between the predicted at least one user memory feature and the one or more memory features stored in the memory node to provide a plurality of bond factors; determining that none of the plurality of bond factors satisfies a threshold; and in response to determining that none of the plurality of bond factors satisfies the threshold: updating the memory data structure by inserting a new memory node storing the predicted at least one user memory feature in the memory data structure; and generating, using the updated memory data structure, an interactive communication for execution by the AIIC. ||
19. A non-transitory computer-readable medium storing software code that, when executed by a hardware processor, causes the hardware processor to: elicit, via an artificial intelligence interactive character (AIIC), a reminiscence from a user; predict, using a trained machine learning (ML) model, at least one user memory feature from the reminiscence; compute, for each memory node of the plurality of memory nodes, a bond factor between the predicted at least one user memory feature and the one or more memory features stored in the memory node to provide a plurality of bond factors; determine that none of the plurality of bond factors satisfies a threshold; and in response to determining that none of the plurality of bond factors satisfies the threshold: update the memory data structure by inserting a new memory node storing the predicted at least one user memory feature in the memory data structure; and generate, using the updated memory data structure, an interactive communication for execution by the AIIC.